Self-Adaptive, Deadline-Aware Resource Control in Cloud Computing

Modern data centers deliver resources over the cloud for clients to run various applications and jobs with diverse requirements. Today's cloud resource management is able to support certain Quality of Service (QoS) requirements including reliability and security. However, in many settings such as the military cloud where latency requirement is paramount, existing cloud resource management schemes fall short in providing a systematic framework to meet and balance disparate types of application deadlines, since they are primarily focused on speeding up job executions for timely processing. In this paper we present a self-adaptive, deadline-aware resource control framework that can be implemented in a fully distributed fashion, making it suitable for unreliable environments where a single point of failure is not acceptable. Relying on Nash Bargaining in non-cooperative game theory, our framework allocates cloud resources in an optimal way to maximize the Nash Bargaining Solutions (NBS) with respect to both job priority and deadline. Further, it also enables self-adaptive deadline-aware resource allocation and rebalancing under cyber or physical attacks that may diminish cloud capacity. We validate our technique by performing experiments on the Hadoop framework.